Military simulation physical engine and separation speed calculation method based on deep learning algorithm

Abstract

The invention discloses a military simulation physical engine and a separation speed calculation method based on a deep learning algorithm. The military simulation physical engine comprises an acting force generator, a rigid body object simulator, a collision detector and a collision processor, the acting force generator is used for detecting the current state of a military simulation program and calculating the acting force required by an object; the rigid body object simulator is used for processing the motion behavior of the rigid body when the acting force is fed back; the collision detector is used for determining a collision result and a contact behavior between objects or between the object and a static scene; a collision detector generates a collision set for use by a collision processor; and the collision processor is used for processing the collision set and adjusting the motion behavior of the rigid body object so as to accurately describe the collision effect.

Description

Technical Field

[0001] The present invention relates to a physics engine and a calculation method thereof, and in particular to a military simulation physics engine and a separation speed calculation method thereof based on a deep learning algorithm. Background Art

[0002] A physics engine is a tool for calculating and simulating physical phenomena, and is widely used in video games, computer animation, virtual reality, scientific research, and other fields. According to different application requirements, physics engines can be divided into real-time engines and high-precision engines. Real-time engines are mainly used for interactive computing, such as games and simulations, and are characterized by fast computing speed but relatively low accuracy; high-precision engines are used in scientific research, computer animation, and other fields, and require higher processing power to obtain more accurate physical effects.

[0003] The current status of physics engine research at home and abroad is as follows:

[0004] 1. Current situation abroad:

[0005] Foreign countries started early in the research of physical engines, and have a history of decades of development. For example, the SRI Research Center in the United States began to study the use of VR technology for military aircraft and vehicle driving training in 1991. In addition, the Department of Computer Science at the University of North Carolina in the United States is also a leader in VR research. They successfully developed the first Walk-through virtual building roaming system for architectural design.

[0006] 2. Domestic situation:

[0007] In recent years, my country has also made significant progress in the research of physical engines. Many domestic scientific research institutions and universities are committed to the development and application of physical engines, covering fields such as virtual reality, games, and animation. For example, some universities and research institutions in my country are conducting research on the application of virtual reality technology in the fields of medicine, aerospace, and transportation planning. In the industrial sector, some companies have also begun to get involved in the development of physical engines, providing technical support for industries such as virtual reality and games.

[0008] In the future, with the continuous advancement of technology, physics engines will play an important role in more fields and bring more rich and diverse interactive experiences to humans.

[0009] Physics engine technology has achieved remarkable results in the fields of games, animation, virtual reality, etc., but there are still some defects and challenges. The following are some defects of existing physics engine technology:

[0010] 1. Performance optimization: Although physics engines have made great progress in computing power and graphics rendering, there are still problems with performance optimization. When dealing with a large number of object collisions, friction, and dynamic effects, physics engines may become too complex and consume performance. Therefore, developers need to constantly seek performance optimization strategies, such as simplifying object models, using collision detection algorithms, etc.

[0011] 2. Real-time performance: There is still room for improvement in the real-time performance of physics engines. In some application scenarios, such as racing games, action-adventure games, etc., real-time calculation of physical effects is crucial to the smoothness and authenticity of the game. However, when dealing with complex physical problems, existing physics engines may experience certain delays, affecting the gaming experience.

[0012] 3. Interactivity: Although the physics engine can simulate the interaction between objects, its performance is still limited in complex scenes, such as special effects such as fragmentation, viscosity, and fluid. In addition, the integration of the physics engine with other technical fields such as game engines and animation engines also needs to be further optimized to achieve more natural and realistic interactive effects.

[0013] 4. Ease of use: The programming interface of the physics engine may be complicated for developers, especially for beginners. Improving the ease of use of the physics engine and lowering the development threshold is an issue worthy of attention.

[0014] 5. Cross-platform compatibility: With the diversification of games and application scenarios, the physics engine needs to adapt to the needs of different platforms (such as PC, mobile devices, cloud games, etc.). It is an important task to achieve cross-platform compatibility while maintaining performance and effects.

[0015] 6. Intelligence: There is still a lot of room for the development of the physics engine in terms of intelligence. For example, by introducing artificial intelligence technology, the physics engine can better adapt to changes in game scenes and automatically adjust parameters to achieve optimal performance.

[0016] 7. Energy saving and emission reduction: With the continuous improvement of green environmental awareness, energy saving and emission reduction has become an important direction for the development of physics engine technology. Under the premise of ensuring the effect, reducing the energy consumption and resource occupation of the physics engine is an urgent problem that needs to be solved.

[0017] In short, although the existing physics engine technology has achieved remarkable results in many aspects, there are still some defects and challenges. In order to better meet the ever-changing application needs, the future physics engine technology needs to continue to develop and improve in terms of performance, real-time, interactivity, ease of use, cross-platform compatibility, intelligence, and energy saving and emission reduction. Summary of the invention

[0018] In view of the highly interactive characteristics of land combat environments, the present invention proposes a military simulation physics engine based on deep learning, focusing on solving the problem that the current combat simulation physics engine has insufficient precision when processing rigid body collisions.

[0019] A military simulation physics engine includes a force generator, a rigid object simulator, a collision detector, and a collision processor; its characteristics are:

[0020] The force generator is used to detect the current state of the military simulation program and calculate the force required by the object;

[0021] The rigid body object simulator is used to process the motion behavior of the rigid body when feeding back the above-mentioned force;

[0022] The collision detector is used to determine the collision results and the contact behavior between objects or between objects and static scenes; the collision detector generates a collision set for use by the collision processor;

[0023] The collision processor is used to process the collision set and adjust the motion behavior of the rigid body object, thereby accurately describing the collision effect.

[0024] The present invention also discloses a separation speed calculation method based on a deep learning algorithm, which is characterized by: when the above-mentioned military simulation physics engine processes a rigid body collision, the calculation method steps are executed:

[0025] (1) Calculate the coordinate set relative to the current collision;

[0026] (2) Calculate the velocity change at the collision point of the object for each unit impulse. Since impulse can cause linear motion and angular velocity, this value needs to consider the linear velocity and angular velocity components along each axis of the coordinate system;

[0027] (3) For the desired speed change, the final result can be converted and the required impulse can be calculated to generate any speed change. Here, a deep neural network is used to generate speed changes.

[0028] Beneficial Effects

[0029] Aiming at the demand for fine entity interaction during land combat simulation, the present invention proposes a physical engine design method that can simulate multiple mechanical calculations and support separation velocity calculation based on deep learning, which can effectively improve the precision of entity mechanical calculations and collision processing in land combat simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a composition diagram of the military simulation physics engine based on deep learning of the present invention;

[0031] Figure 2 It is a schematic diagram of the structure of the force generator of the present invention;

[0032] Figure 3 It is a schematic diagram of the structure of the rigid object simulator of the present invention;

[0033] Figure 4 It is a schematic diagram of intersection depth processing of the present invention;

[0034] Figure 5 This is a schematic diagram of edge-to-edge collision intersection depth processing of the present invention;

[0035] Figure 6 This is a depth processing diagram of edge-face collision intersection of the present invention;

[0036] Figure 7 This is a depth processing diagram of the face-to-face collision intersection of the present invention;

[0037] Figure 8 This is a schematic diagram of the structure of the collision processor of the present invention;

[0038] Fig. 9 A schematic diagram for collision generation of the present invention;

[0039] Fig.10 Deep neural network graph for intersection velocity processing of the present invention. DETAILED DESCRIPTION

[0040] Example 1

[0041] See also Figure 1 As shown, the present invention discloses a military simulation physics engine, including a force generator, a rigid object simulator, a collision detector, and a collision processor; wherein:

[0042] See attached Figure 2 As shown. Force generator: used to detect the current state of the military simulation program and calculate the force required by the object; the force generator consists of two parts: particle physics module and mass aggregate force accumulator.

[0043] (1) Particle physics module: used to simulate fireworks and display explosion, water flow, smoke and flame effects; construct the initial physics engine, perform vector mathematical operations and the laws of motion they support, and complete the physical calculation of the ballistic trajectories of various launch objects (bullets, shells, fireballs and laser beams). The basic physics formula is as follows:

[0044] e=ms 2

[0045] Among them, e represents the energy value, m represents the mass of the launching object, and s represents the velocity of the launching object. If the energy conservation is maintained, the mass change can be calculated based on the known velocity change.

[0046] Δm=(Δs) 2

[0047] Δm represents the change in mass of the launched object, and Δs represents the change in velocity of the launched object;

[0048] (2) Mass aggregate force accumulator

[0049] The mass aggregate force accumulator connects multiple masses through springs and connecting rods, and can simulate any object based on the particle physics system.

[0050] D'Alembert's theorem states that if there is a set of forces acting on an object, then all of them can be replaced by a single force, as shown below:

[0051]

[0052] f i represents the force acting on the object in the i-th direction, and f represents the total force replaced by D'Alembert's theorem;

[0053] Use a vector as a force accumulator. In each frame, reset the vector to zero and add each force in turn through vector addition.

[0054] Some forces result from the motion of different objects, such as drag. Some forces result from the interaction between objects and their environment, such as buoyancy and explosive forces. Some forces result from the way objects are connected.

[0055] The force generator is based on the data processed by the particle physics module and the mass aggregate force accumulator. It performs cumulative vector calculations on the numerical calculation results of particle simulation mechanics and the numerical calculation results of mass aggregate cumulative mechanics, and increases the force it receives in the rigid body class.

[0056] Rigid body object simulator: used to process the motion behavior of the rigid body when the above force feedback is applied. The rigid body object simulator includes the gravity simulator, the resistance simulator, and the elastic force simulator. Figure 3 shown.

[0057] (1) Gravity generation simulator

[0058] The gravity simulator is implemented by particle binding, which is achieved by multiplying the particle mass by the gravity coefficient and adding a force accumulator.

[0059] G p =m p g

[0060] Among them, G p is the gravitational force on the particle, m p is the particle mass, g is the gravity coefficient;

[0061] (2) The resistance generation simulator adopts the following simplified resistance model:

[0062]

[0063] Among them, k1 and k2 represent the resistance constants, represents the normalized velocity of the particle, f drag Represents the drag being simulated and depends on the object and drag type.

[0064] (3) The elastic force simulator is mainly used to simulate the suspension system of tracked wheeled equipment. Its theoretical basis is Hooke's law:

[0065] f e =-kΔl

[0066] Where Δl represents the distance the spring is extended or compressed, k represents the elastic coefficient, and f e Represents the simulated elastic force, which reflects the hardness of the spring:

[0067] f e =-k(l-l0)

[0068] When simulating buoyancy, the force is calculated according to the following formula;

[0069] f f =m o g

[0070] Among them, f f is the simulated buoyancy, m o It is expressed as mass and g is the acceleration due to gravity.

[0071] The rigid body object simulator performs cumulative calculations on the mechanical vector values ​​generated by the gravity simulator, the resistance simulator, and the elastic force simulator, and synchronously updates the calculated mechanical vectors in the rigid body object class.

[0072] Collision detector: used to determine the collision results and contact behavior between objects or between objects and static scenes. The collision detector generates a collision set for use by the collision processor;

[0073] The collision detector is responsible for handling the interaction process generated by the collision between rigid bodies. It calculates the set of objects that may cause collision (rough collision detection), checks the candidate collision results, and then completes the collision generation process to provide mechanical feedback to the rigid body.

[0074] The first stage of collision detection should generate a detailed content detection table. The bounding volume can be regarded as a spatial volume that surrounds an object. It uses a relatively simple geometric shape, usually a sphere or a box.

[0075] For two spheres, it is convenient to check whether the two are in an overlapping state. Specifically, if the distance between the centers of the spheres is less than the sum of their radii, the two spheres are determined to intersect with each other.

[0076] Cubic boxes are also commonly used as bounding volumes and can be described by a center point and a set of dimensions.

[0077] There are two common application methods for box bounding volumes, namely the world coordinate axis-aligned box (axis-aligned bounding box or AABB) and the object coordinate axis-aligned box (object bounding box or OBB). The bounding volume hierarchy (BVH) is used to contain the bounding volume at the leaf node of the tree structure and covers each operation object. The bottom-level bounding volume is connected to its parent node, and the latter also contains its own bounding volume; if two high-level nodes intersect, it is necessary to further examine the child nodes of each node. For different types of collisions, there are the following solutions:

[0078] 1) Point-surface collision

[0079] Point-to-plane collision is common and can be considered an important type of collision. The collision normal is determined by the surface normal at the collision point. If the object point passes through the current surface, it is usually necessary to project the point back onto the surface to calculate the collision position.

[0080] 2) Edge-to-edge collision

[0081] Edge-edge collision is also a relatively common type of collision (second only to point-plane collision), and can be regarded as a key factor in static collisions between objects and planes or curved surfaces.

[0082] The collision normal is perpendicular to the tangents of the two edges, and the intersection depth is expressed as the distance between the two edges.

[0083] 3) Edge-face collision

[0084] Edge-face collisions are only used for curved surfaces (such as corresponding edges in a capsule).

[0085] 4) Face-to-face collision

[0086] Face-to-face collisions arise from the collision between curved surfaces and other surfaces, such as a sphere on a flat surface, and their collision data is somewhat arbitrary compared to other collision behaviors.

[0087] When a rotating object collides with the ground, in addition to bouncing upward, the object's angular velocity typically changes.

[0088] The forces acting on an object will produce linear acceleration and angular velocity, where the linear acceleration term is as follows:

[0089]

[0090] The angular velocity term based on the torque is as follows:

[0091] τ=p f ×f

[0092] Here, torque generates angular acceleration as follows:

[0093]

[0094] 5) Rotational collision

[0095] If the two collision points are tracked at the time of collision, their separation speed is as follows:

[0096] v′ s =-cv s

[0097] Among them, v s represents the relative velocity between objects before collision, v′ s represents the relative velocity after the collision, and c represents the coefficient of rebound.

[0098] When dealing with the relative motion between two objects, the following four impulses need to be calculated: the linear impulse and the angular impulse of the two objects. If only a single object is involved in the collision (for example, an object collides with a stationary object such as the ground), only the impulse and impact torque of the single object need to be calculated, and the following steps need to be performed:

[0099] A) Calculate the coordinate system geometry relative to the collision point, construct a transformation matrix, and then perform transformation operations between the above coordinate system sets;

[0100] B) Calculate the velocity change per unit impulse at each object's collision point. Since impulse causes both linear motion and angular velocity of the object, this value needs to consider both separately;

[0101] 1) Linear component:

[0102] The processing of the linear component is relatively simple, based on the fact that the linear velocity change of each unit impulse is consistent with the impulse direction, that is, the inverse of the mass, as shown below:

[0103]

[0104] is the inverse of the mass, and m is the mass of the collider.

[0105] For a collision between two objects, the linear component is simply expressed as the sum of two inverse masses, as follows:

[0106]

[0107] is the inverse of the collision mass between two objects, is the inverse of the mass of collision body a, It is the inverse of the mass of collision body b.

[0108] 2) Angular velocity component

[0109] q rel =qp

[0110] q rel is the collision position vector, which is used to record the collision position relative to the collision origin. q is the collision position coordinate, and p is the collision origin coordinate. The impact torque from a unit impulse can be expressed as:

[0111]

[0112] Among them, u is the impact torque, q rel is the collision position vector, d is the impulse direction, and the angular velocity change based on the unit impact torque is as follows:

[0113]

[0114] in, is the change in angular velocity caused by the impulse, I is the collision position coordinate transformation matrix, and u is the impact torque.

[0115] Rotation speed:

[0116] in, To calculate the rotation speed, is the angular velocity, q rel is the collision position vector;

[0117] C) calculating the separation velocity at the collision point, the current closing velocity, and the difference between the two, the result of which is the expected velocity change;

[0118] D) Based on the expected speed change, the impulse generated can be calculated;

[0119]

[0120] v represents the desired velocity change, g represents the required impulse, and d represents the intersection distance.

[0121] First, the closing speed at the collision point needs to be calculated, and then the speed change value needs to be calculated. The speed includes linear and angular components. For the overall speed of an object at the collision point, the two must be calculated separately. Accordingly, the linear speed and the linear speed at the collision point based on rotation can be calculated.

[0122] v′ s =-cvs

[0123] Δv s =-v s -cv s =-(1+c)v s

[0124] If the spring factor is 0, the change in velocity is enough to remove all existing closing velocity; if the spring factor is close to 1, the objects separate at approximately the same speed as when they approached.

[0125] Collision Processor: It is used to process collision sets and adjust the motion behavior of rigid body objects to accurately describe the collision effect. When performing rigid body dynamics simulation through a rigid body simulator, according to the above collision detection process, if a collision occurs, the collision processing is performed according to the following steps.

[0126] Collision handling solution: Closing velocity refers to the combined velocity of two objects when they move together. When two objects collide, they compress each other, and the force generated by the elastic deformation of their surfaces separates the objects. Assuming that momentum is conserved during the collision, as shown below:

[0127]

[0128] Among them, m a Expressed as the mass of object a, It is represented as the velocity of object a before collision, It is expressed as the speed after the collision; m b Expressed as the mass of object b, Represented as the velocity of object b before the collision, is the velocity after the collision.

[0129] The independent velocities can be related using closing velocities:

[0130] v′ s =-cv s

[0131] Among them, v′ s Expressed as the separation velocity after collision, v s is expressed as the separation velocity before the collision, and the constant c is expressed as the coefficient of rebound.

[0132] If two objects are partially embedded in each other, they are said to be in an intersecting state.

[0133] When two objects are in an intersecting state, they must be separated, and the total moving distance of the objects is equal to the depth value when the two objects intersect, as shown below.

[0134] Δp a +Δpb =d

[0135] Among them, Δp a Expressed as the scalar distance moved by object a, Δp b It is represented as the scalar distance moved by object b, d represents the total distance moved by the object, and the two distances are related according to the ratio of the object masses as follows:

[0136] m a Δp a =m b Δp b

[0137] After integration, the following formula is generated:

[0138]

[0139]

[0140]

[0141]

[0142] Among them, Δp a Expressed as the scalar distance moved by object a, Δp b It is represented as the scalar distance moved by object b, d represents the total movement distance of the object, and n represents the collision normal.

[0143] 1) Static collision

[0144] The particle is at rest on the ground and is affected by a single force, gravity. The particle moves downward and begins to intersect the ground, which is picked up by the collision detector and generates a collision message.

[0145] The collision solver looks at the particles and sees if they have the following intersection velocities:

[0146]

[0147] After collision feedback, the particle's velocity looks like this:

[0148]

[0149] The collision processor receives the collision table from the collision detection system and updates the simulation objects to examine the relevant collision behavior.

[0150] Time Division Engine: If there is no collision, the object can move freely according to the above laws of motion and the force generator.

[0151] When a collision occurs, the two objects touch each other at a certain point, but they do not intersect.

[0152] If there are multiple collisions, they can be processed in sequence, and between collisions the current scene world can be updated using the laws of motion.

[0153] The following algorithm steps are used:

[0154] (1) Let the initial time be the current simulation time and the end time be the end time of the current update operation, that is:

[0155] t0=t s

[0156] t1=t s+1

[0157] Among them, t0 is the initial time, t s is the current simulation time, t1 is the end time, t s+1 The end time of the current update operation.

[0158] (2) Perform update operations for the simulation step time interval, as follows:

[0159] The time interval can be recorded as: Δt = t1-t0

[0160] The simulation step length is t step , then the number of updated tables in this time interval can be calculated as follows:

[0161]

[0162] n step This is the number of collision tables.

[0163] (3) Run the collision detector according to the above collision detection scheme, collect and store the collision data in the collision table.

[0164] (4) If there is no collision, exit the current algorithm.

[0165] (5) Read the collision table contents one by one and calculate the accurate first collision time for each collision, recorded as t n1

[0166] (6) Select the first collision behavior and record it as event1.

[0167] (7) If the first collision occurs after the termination time, exit the current algorithm.

[0168] (8) Remove all update operations from step (2) and perform update operations based on the start time of the first collision.

[0169] (9) Process the current collision behavior through impulse calculation.

[0170] (10) Set the initial time to the first collision time, keep the end time unchanged and return to step (1).

[0171] Example 2

[0172] The military simulation physics engine designed by the present invention proposes a separation speed calculation method based on a deep learning algorithm when processing rigid body collisions, as detailed below:

[0173] When rigid bodies collide, an impulse is generated in a certain direction and the collision speed in different directions is increased. The processing algorithm includes the following steps: - The following steps need to be combined with the above-mentioned "engine" content to explain in detail how each step is implemented, otherwise it will not be fully disclosed.

[0174] (1) According to the collision information stored in the aforementioned collision table, read and calculate the coordinate set relative to the current collision;

[0175] (2) Calculate the velocity change at the collision point of the object for each unit impulse. Since impulse can cause linear motion and angular velocity, this value needs to consider the linear velocity and angular velocity components along each axis of the coordinate system;

[0176] The change in line speed can be calculated as follows:

[0177]

[0178] is the rigid body velocity change, g r is the impulse data, m r is the rigid body mass.

[0179] The rotation change is calculated using the following formula:

[0180]

[0181] is the rotation change, I is the rotation angle,

[0182] The impact torque u needs to be calculated according to the following formula:

[0183] u=q rel ×g

[0184] u is the impact torque, q rel is the direction vector, and g is the impulse change value.

[0185] (3) For the desired speed change, the final result can be converted and the required impulse can be calculated to generate any speed change. Here, a deep neural network is used to generate speed changes.

[0186] The specific process is as follows:

[0187] The network input is the sequential collision events in the collision set, including the 3D image and rigid body data information (position, posture, etc.) of the current collision event;

[0188] Convolutional neural network is used to extract features from 3D images, and the features are fitted with 3D data information and then enter the feature processing layer for further extraction.

[0189] The extracted data information is transmitted to the fully connected layer for nonlinear fitting, and the material data is subjected to nonlinear fitting through the fully connected layer to obtain the separated linear velocity components and angular velocity components of the collision objects A and B in the x, y, and z axes respectively;

[0190] Based on deep learning technology and physical mechanics theory, the present invention integrates key technical elements in the professional field of combat simulation and designs a professional physics engine for physical interactive computing simulation of combat simulation. Based on the present invention, the "entity-environment" high-precision interactive simulation function can be realized, and the separation speed calculation is optimized and designed by deep learning technology, which can provide refined physical interactive computing for land combat simulation deduction.

[0191] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A military simulation physics engine, comprising a force generator, a rigid object simulator, a collision detector, and a collision processor; its characteristics are: The force generator is used to detect the current state of the military simulation program and calculate the force required by the object; The rigid body object simulator is used to process the motion behavior of the rigid body when feeding back the above-mentioned force; The collision detector is used to determine the collision result and the contact behavior between objects or between objects and static scenes; The collision detector generates a collision set for use by the collision processor; The collision processor is used to process the collision set and adjust the motion behavior of the rigid body object, thereby accurately describing the collision effect.

2. The military simulation physics engine according to claim 1, characterized in that: The force generator comprises two parts: a particle physics module and a mass aggregate force accumulator; Particle physics module: used to construct the initial physics engine, perform vector mathematical operations and the laws of motion they support, and complete the physical calculation of the ballistic trajectories of various launch objects. The basic physics formulas are as follows: e=ms 2 Among them, e represents the energy value, m represents the mass of the emitted object, and s represents the velocity of the emitted object. If the energy conservation is maintained, the mass change can be calculated based on the known velocity change: Δm=(Δs) 2 ; Δm represents the change in mass of the launched object, and Δs represents the change in velocity of the launched object; Mass Aggregate Force Accumulator: Connect multiple masses through springs and connecting rods to simulate any object based on particle physics system: D'Alembert's theorem states that if there is a set of forces acting on an object, then all of them can be replaced by a single force, as shown below: f i represents the force acting on the object in the i-th direction, f represents all the forces replaced by D'Alembert's theorem, in each frame, the vector is reset to zero, and each force is added in turn by vector addition.

3. The military simulation physics engine according to claim 1, characterized in that: The rigid object simulator includes a gravity simulator, a resistance simulator, and an elastic force simulator: (1) The gravity simulator is implemented by particle binding, which is achieved by multiplying the particle mass by the gravity coefficient and adding a force accumulator: G p =m p g Among them, G p is the gravitational force on the particle, m p is the particle mass, g is the gravity coefficient; (2) The resistance generation simulator adopts the following simplified resistance model: Among them, k1 and k2 represent the resistance constants, represents the normalized velocity of the particle, f drag represents the simulated resistance and depends on the object and the resistance type; (3) The elastic force simulator is used to simulate the suspension system of tracked wheeled equipment. Its theoretical basis is Hooke's law: in e =-kΔl Where Δl represents the distance the spring is extended or compressed, k represents the elastic coefficient, and f e Represents the simulated elastic force, which reflects the hardness of the spring: f e =-k(l-l0) When simulating buoyancy, the force is calculated according to the following formula; f f =m o g Among them, f f is the simulated buoyancy, m o It is expressed as mass and g is the acceleration due to gravity.

4. The military simulation physics engine according to claim 1, characterized in that: The collision detector is responsible for processing the interaction process generated by the collision between rigid bodies. It calculates the set of objects that may cause collision, checks the candidate collision results, and then completes the collision generation process to provide mechanical feedback to the rigid body.

5. The military simulation physics engine according to claim 1, characterized in that: The collision handling solution is mainly used to handle collision sets and adjust the motion behavior of rigid objects to accurately describe the collision effect: the closing velocity refers to the combined velocity of two objects when they move together. When two objects collide, they compress each other, and the force generated by the elastic deformation of their surfaces separates the objects. Assuming that the collision process maintains momentum conservation, it is shown below: Among them, m a Expressed as the mass of object a, It is represented as the velocity of object a before collision, It is expressed as the speed after the collision; m b Expressed as the mass of object b, Represented as the velocity of object b before the collision, It is expressed as the velocity after the collision; The independent velocities can be related using closing velocities: v′ s =-cv s Among them, v′ s Expressed as the separation velocity after collision, v s Then is expressed as the separation velocity before collision, and the constant c is expressed as the coefficient of rebound; If two objects are partially embedded in each other, they are said to be in an intersecting state; When two objects are in an intersecting state, they must be separated, and the total moving distance of the objects is equal to the depth value when the two objects intersect, as shown below: Δp a +Δp b =d Among them, Δp a Expressed as the scalar distance moved by object a, Δp b It is represented as the scalar distance moved by object b, d represents the total distance moved by the object, and the two distances are related according to the ratio of the object masses as follows: m a Δp a =m b Δp b After integration, the following formula is generated: Among them, Δp a Expressed as the scalar distance moved by object a, Δp b It is represented as the scalar distance moved by object b, d represents the total movement distance of the object, and n represents the collision normal; 1) Static collision The particle is stationary on the ground and is affected by a single force, gravity. The particle moves downward and begins to intersect the ground, which is captured by the collision detector and generates collision information. The collision solver looks at the particles and sees if they have the following intersection velocities: After collision feedback, the particle's velocity looks like this: The collision processor receives the collision table from the collision detection system and updates the simulation object to examine the relevant collision behavior; Time Division Engine: If there is no collision, the object can move freely according to the aforementioned laws of motion and the force generator; When a collision occurs, the two objects touch each other at a certain point, but they do not intersect; If there are multiple collisions, they can be processed in sequence, and between collisions the current scene world can be updated using the laws of motion.

6. The military simulation physics engine according to claim 5, characterized in that: The collision handling scheme uses the following algorithmic steps: (1) Let the initial time be the current simulation time and the end time be the end time of the current update operation; (2) performing an update operation for the simulation step time interval; (3) Run the collision detector and collect the collision table; (4) If there is no collision, exit the current algorithm; (5) Calculate the accurate first collision time for each collision; (6) Select the first collision behavior; (7) If the first collision occurs after the termination time, exit the current algorithm; (8) Remove all update operations from step (2) and perform update operations based on the start time of the first collision; (9) Process the current collision behavior through impulse calculation; (10) Set the initial time to the first collision time, keep the end time unchanged and return to step (1).

7. A separation speed calculation method based on a deep learning algorithm, characterized by: When the military simulation physics engine described in any one of claims 1 to 6 processes a rigid body collision, the calculation method steps are executed: (1) Calculate the coordinate set relative to the current collision; (2) Calculate the velocity change at the collision point of the object for each unit impulse. Since impulse can cause linear motion and angular velocity, this value needs to consider the linear velocity and angular velocity components along each axis of the coordinate system; (3) For the desired speed change, the final result can be converted and the required impulse can be calculated to generate any speed change. Here, a deep neural network is used to generate speed changes.

8. A non-volatile storage medium, characterized by: The non-volatile storage medium includes a stored program, wherein when the program is executed, the device where the non-volatile storage medium is located is controlled to execute the method according to claim 7.

9. An electronic device, characterized by: It comprises a processor and a memory; the memory stores computer-readable instructions, and the processor is used to execute the computer-readable instructions, wherein the computer-readable instructions execute the method according to claim 7 when they are executed.

Images